plm, controls, model, effect and index, fixed effects and random effects

[LucasGomes]

I have MANY doubts when TRYING to use the plm() function.
Let's say my data.frame "df" has many variables as columns, e.g. ROS, type of business/industry, employees, assets, etc. It has these values for many businesses, and for each year.
For simplicity's sake, it could look like something like this:

> df
Year  ID     Type     ROS       Employees     etc...
2010   1       55     103            4         ...
2011   1       55     120            6         ...
2012   1       55     111            6         ...
2010   2       42     106            5
2011   2       42     195           20
2012   2       42     214           20
2010   3       55     130            9
2011   3       55      95           12
2012   3       55      40            2

1. When using plm is there a difference where we index the ID and time variables? In the pdata.frame or plm functions? I.e. should I do this:

panel <- pdata.frame(df, index = c("ID", "Year"))

and then not include the index in the plm function or should I put it directly into the plm function?

2. How do I include "assets" as a control variable for "ROS" as dependent and "Employees" as independent?
   Would it be this?

plm(ROS ~ Employees + Assets, index = c("ID", "Year"), 
model = "within", effect = "twoways", data = panel)

3. What is the difference between the fixed effects or random effects models?

4. If I want the "ID" and "Year" variables to be fixed should I use the "twoways" effect or the "nested"? What is the difference between them?

5. How can I add a third variable to the index? The "Type" variable, for example. Does this work?

plm(ROS ~ Employees, index = c("ID", "Year", "Type"), model = "within", 
effect = "twoways", data = df)

Or does the factor() function do the same thing? Could I even use the "twoways" effect in this? Is there a "threeways" effect?

plm(ROS ~ Employees + factor(Type), index = c("ID", "Year"), 
model = "within", effect = "twoways", data = df)

Would doing this previous line of code give me the "Employees" estimated coefficient I'm looking for?

6. Would using the "random" effects model solve this issue? Something like this:

plm(ROS ~ Employees + factor(Type), index = c("ID", "Year"), 
model = "random", effect = "twoways", data = df)

Does using the "twoways" effect work for a model = "random"?

7. How would I estimate "Type" and "Year" fixed effects and "ID" random effects using plm()?